1. law, govt and politics
2. immigration

Marc Piopiunik and Jens Ruhose:
Immigration, Regional Conditions, and Crime:
Evidence from an Allocation Policy in Germany
Munich Discussion Paper No. 2015-3
Department of Economics
University of Munich
Volkswirtschaftliche Fakultät
Ludwig-Maximilians-Universität München
Online at http://epub.ub.uni-muenchen.de/24468/
Immigration, Regional Conditions, and Crime:
Evidence from an Allocation Policy in Germany
Marc Piopiunik∗
Jens Ruhose†
March 24, 2015
Abstract
After the collapse of the Soviet Union, more than 3 million people with
German ancestors immigrated to Germany under a special law granting immediate
citizenship. Exploiting the exogenous allocation of ethnic German immigrants
by German authorities across regions upon arrival, we find that immigration
significantly increases crime. The crime impact of immigration depends strongly
on local labor market conditions, with strong impacts in regions with high
unemployment. Similarly, we find substantially stronger effects in regions with
high preexisting crime levels or large shares of foreigners.
JEL-Code: F22, J15, K42, R10
Keywords: Immigration, crime, allocation policy
We are grateful to George J. Borjas, Oliver Falck, Albrecht Glitz, Eric A. Hanushek, Ludger
Woessmann, and participants at the Ifo Institute in Munich, ESPE in Braga, EEA in Toulouse, and
EALE in Ljubljiana for helpful comments and discussion. This project was funded through the Leibniz
Competition (SAW-2012-ifo-3).
∗
Ifo Institute - Leibniz Institute for Economic Research at the University of Munich, Poschingerstr.
5, 81679 Munich, Germany and CESifo; E-mail: [email protected]; Phone: +49 89 9224 1312
†
Ifo Institute - Leibniz Institute for Economic Research at the University of Munich, Poschingerstr.
5, 81679 Munich, Germany and IZA; E-mail: [email protected]; Phone: +49 89 9224 1388
1
Introduction
Criminal behavior of immigrants is a huge concern in many countries. In Europe, the
majority of native residents is worried about immigrants increasing crime, whereas only
a minority is worried about immigrants taking jobs away (Fitzgerald et al., 2012). The
widespread concern about crime might therefore play a greater role in shaping immigration
policies than do labor market concerns (Card et al., 2012). While the literature on the
labor market effects of immigration is huge (see Friedberg and Hunt, 1995; Borjas, 1999;
Card, 2009, for overviews), the crime effects of immigration has been explored much less.
Existing research on the crime impact of immigration tends to find zero or small
effects. Bianchi et al. (2012) exploit the increase in the immigrant population in Italy
in the 1990s that was mainly driven by the collapse of the Soviet Union and the Balkan
Wars, finding strong effects on robberies, but a negligible effect on total crime. Bell
et al. (2013) study the impact on crime of two different immigration waves to the United
Kingdom: migrant laborers from eight European countries that joined the EU in 2004
and asylum seekers. The authors find an effect only for asylum seekers, and only on
property crime. For the United States, Butcher and Piehl (1998, 2007) find no significant
relationship between immigration and crime at the state level, and Chalfin (2013), using
data at the metropolitan area level, finds no evidence that Mexican immigrants increase
crime. Spenkuch (2013) uses county data and finds effects on property crime, but not on
violent crime.
We contribute to this literature by investigating the crime impact of a particular group
of low-educated immigrants to Germany who were allocated across regions by German
authorities upon arrival. We find that these immigrants increased crime substantially. In
particular, crime effects are much stronger in regions with adverse conditions such as high
unemployment and high preexisting crime levels.
The group of immigrants we focus on in this study are ethnic Germans who lived in
Eastern Europe and the former Soviet Union before migrating to Germany after the fall of
the Berlin Wall. Between 1988 and 2005, more than 3 million ethnic Germans immigrated,
increasing Germany’s population by about 5%. Immediately after immigrating, ethnic
German immigrants were granted German citizenship, were allowed to work, and were
eligible to social security assistance (like German natives). Despite legally being Germans,
ethnic German immigrants have considerably lower education levels, worse labor market
outcomes, and lower incomes than native Germans (see Table A-2). Based on these
characteristics, however, ethnic German immigrants are similar to other immigrants in
Germany.
Upon arrival in Germany, ethnic German immigrants are allocated across counties
by state authorities. We exploit this allocation policy to identify the causal effect of
1
ethnic German immigrants on crime.1 The number of migrants allocated to a particular
county is set by policymakers and depends largely on the number of residents, economic
conditions, and other factors such as housing capacities. While German authorities
determine the number of ethnic German immigrants allocated to a particular county,
each immigrant can express preferences to live in the county where her relatives are
already living. If the county quota is not already exhausted, authorities try to meet these
family preferences.2 Therefore, the specific county of residence where a newly arriving
ethnic German immigrant is allocated to, depends mostly on family ties. Noncompliance
with the residence allocation, which is binding for three years, is severely sanctioned with
the withdrawal of all social benefits, implying that self-selection into regions is highly
restricted. Since the residence of family members is the main allocation criterion and labor
market skills do not play any role (Glitz, 2012, p. 180), the allocation is likely exogenous
with respect to regional crime and labor market conditions. This allocation, therefore,
provides a unique quasi-experimental setting for studying the effects of immigration on
crime.
Furthermore, as the allocation policy has likely led to similar skill and age
distributions of ethnic German immigrants across regions, we are able to assess the
importance of regional conditions for the crime impact.3
We combine annual county-specific inflows of ethnic German immigrants with annual
county-specific crime rates.
We merge information on labor market conditions and
demographics of the counties to investigate whether regional conditions influence the
impact of immigration on crime. We focus on West Germany (excluding Berlin) from
1996 until 2005, the period during which newly arriving ethnic German immigrants were
allocated across regions by German authorities. Thus, our sample includes 185 counties
over a 10-year period.
The results indicate that ethnic German immigrants increase crime rates substantially.
An immigrant inflow of one ethnic German per 1,000 inhabitants increases total crime by
about 0.9%, which amounts to an elasticity of approximately 0.5. Effects vary by type of
crime, with strong impacts on burglary, property damage, and battery, but no effect on
street-related types of crime. We also find evidence that regional labor market conditions
are crucial. While immigration has no effect on crime in regions with low unemployment,
1
Glitz (2012) exploits the same allocation policy to estimate the impact of immigrants on skill-specific
employment rates and wages of natives, finding a displacement but no wage effect. Other studies
exploiting allocation policies (typically involving refugees) include, for example, Edin et al. (2003); Damm
(2009); Damm and Dustmann (2014).
2
Note that although the majority of individual residence choices depended on family ties, the variation
in the size of immigrant flows across regions is not based on networks as in the shift-share approach
(Card, 2001). Indeed, in our case, the regional distribution of resident immigrants cannot be used as an
instrument for the allocation of new immigrants across regions.
3
Hjalmarsson and Lindquist (2012) find that children of criminal fathers are much more likely of having
a criminal conviction than those with noncriminal fathers. We argue in Section 4 that intergenerational
correlation in crime is likely not an issue in our setting.
2
crime effects are strong in regions with high unemployment. Furthermore, we find much
stronger impacts in regions with high preexisting crime levels.
In their economic theory of crime, Becker (1968) and Ehrlich (1973) argue that
the propensity for committing a crime decreases with legitimate earnings opportunities,
the probability of being convicted, and the cost of conviction. The cost of conviction
is particularly interesting since it might differ between immigration groups. Because
citizenship reduces the cost of conviction by eliminating the threat of deportation
(Spenkuch, 2013), this might be one explanation why the crime effects we find are
larger than those found in other studies, given that ethnic Germans are granted German
citizenship immediately upon arrival.4
We contribute to the literature by providing the first assessment of the importance
of local labor market conditions for the crime impact of immigration. Existing studies
either investigate the impact of local labor market conditions on crime in general (without
focusing on immigrants) or study the impact of immigrants on crime without considering
the potentially important role of local labor market conditions. Assessing the importance
of local labor market conditions is feasible in this setting since the allocation policy led
to similar skill and age distributions of newly arriving ethnic German immigrants across
counties. In contrast, immigrants studied elsewhere could always choose their region of
residence without any restrictions (with the exception of refugees). Furthermore, we
estimate short-run effects of immigration on crime by exploiting annual immigration
inflows. Based on changes in immigrant stocks over time, existing studies, in contrast,
estimate medium- or long-run effects. If immigrants are more likely to commit crimes in
the first years after arrival, for example, because integration into the new society takes
some time, then we are able to detect crime effects that have been missed in previous
studies.
The remainder of the paper proceeds as follows. Section 2 provides some background
of ethnic German immigrants and describes the allocation process. Section 3 presents
the data, and Section 4 lays out the empirical model. Section 5 reports the main results,
robustness checks, and effect heterogeneities with respect to regional conditions. Section
6 concludes.
2
Ethnic German Immigrants and the Allocation
Policy
Ethnic German immigrants are descendants of German colonists who had migrated to
Russia and other East European countries in the 18th and 19th century (Bade, 1990).
During World War II, many ethnic Germans—considered potential collaborators with
4
In the United States, for example, 816,000 criminal immigrants were removed from the country
between 1998 and 2007 because of a criminal charge or conviction (Camarota and Vaughan, 2009, p. 2).
3
the Nazi regime—were forced to leave their original settlements and to move eastward to
Siberia, Kazakhstan, and to regions in Middle Asia where they lived in special settlements.
During the Cold War, only few ethnic Germans were allowed to emigrate (see Figure 1).
Toward the end of the communist regime, restrictions were alleviated, leading to a massive
emigration wave driven by bad economic and political conditions in the Soviet Union
(Bade and Oltmer, 2003). Since the end of the 1980s, more than 3 million ethnic German
immigrants have arrived in Germany.
There was a heavy inflow of ethnic German
immigrants in the late 1980s and early 1990s, with a peak of 397,000 individuals in
1990. Due to this huge immigration, a yearly quota of about 225,000 individuals was
introduced in 1993 and further reduced to about 100,000 individuals per year in 2000.
Due to these quotas, and because the stock of German descendants in the former Soviet
Union was becoming smaller, the number of immigrating ethnic Germans decreased over
time.5 The collapse of the Soviet Union changed the composition of countries where
ethnic German immigrants were coming from. Between 1950 and 1987, only about
7% of ethnic German immigrants came from countries of the former Soviet Union. In
contrast, since the mid-1990s, almost all ethnic German immigrants have come from
former Soviet Union countries, especially Kazakhstan, Kirgisistan, Usbekistan, Ukraine,
and the Russian Federation (see Figure 1).
The immigration of ethnic Germans was administered centrally by the Federal Office
of Administration (Bundesverwaltungsamt).
First, all ethnic Germans had to apply
for a visa at the German embassy in their country of residence. To be recognized as
an ethnic German, an applicant had to fulfill all requirements of the Federal Refugees
Act. Legally, individuals are considered members of German minorities if they meet the
following three requirements (Peters, 2003). They (1) have to be descendants of at least
one parent or grandparent with German nationality; (2) must have declared “confession to
German ethnicity” before leaving the settlement area, typically by possessing the German
nationality, sending their children to a German school, or through being a member in a
German association; and (3) are able to conduct a simple conversation in German. Due to
assimilation pressures in their settlement areas, ethnic German immigrants had rather low
German language proficiency, however. Therefore, language requirements were kept at a
low level. To maintain the unity of the family, ethnic Germans were allowed to immigrate
together with their spouses and all offsprings, given that the marriage had existed for at
least three years. Ethnic Germans and all relatives migrating with them were granted
German nationality immediately upon arrival.6
All arriving ethnic German immigrants had to pass through a central admission center.
If ethnic German immigrants did not have a job or another source of income—which was
5
Ethnic German immigrants do not include East Germans who moved to West Germany after the fall
of the Berlin Wall.
6
In the remainder of the paper, the term ethnic German immigrants includes both persons of own
German descent as well as their spouses and children.
4
the case for the vast majority (see below)—they were allocated to one of the 16 German
states according to predefined quotas (K¨onigsteiner Schl¨
ussel). These quotas, which have
been adjusted every year, are based on the tax revenues (weight of two thirds) and the
population size (one third) of each state.
Based on the Place of Residence Allocation Act (Wohnortzuweisungsgesetz), which
has been established in 1989 due to the large inflow, most states subsequently allocated
ethnic German immigrants across counties. The goal of the allocation across counties was
a socially acceptable integration of these immigrants. Because ethnic German immigrants
preferred to live close to their relatives, and because noncompliance with the assignment
did not have any consequences at the beginning of the 1990s, many ethnic Germans
left their assigned county and moved to other counties. This led to a strong inflow of
ethnic German immigrants into some regions (Swiaczny and Mammey, 2001). Because
municipalities were responsible for the housing of ethnic German immigrants and for
the payment of social assistance, this implied great financial burdens for municipalities
with high shares of ethnic Germans. Furthermore, high concentrations of ethnic German
immigrants increased the necessity of integration measures and lowered the acceptance
among native Germans (Haug and Sauer, 2006). Seven counties which were particularly
hit by massive inflows of ethnic German immigrants therefore signed the so-called
“Gifhorn Declaration for the Integration of Ethnic German immigrants” in March 1995,
demanding a more solidary distribution of the burdens between states and municipalities
(Nieders¨achsische Landeszentrale f¨
ur Politische Bildung, 2002, p. 10).7 This declaration
led to an important amendment to the Place of Residence Allocation Act that mandated
that ethnic German immigrants were bound to their assigned county during the first three
years after immigration. Most importantly, the amendment introduced severe sanctions:
Ethnic German immigrants would lose all social benefits and any type of public assistance
in case of noncompliance with the allocation decision. Only ethnic Germans who could
prove both a job and housing in their preferred county were exempt from this rule.
However, survey evidence (presented below) indicates that only a minority of arriving
ethnic German immigrants was able to freely choose their county of residence. Most West
German states implemented the new law in March 1996, while Hesse adopted the law in
January 2002, and two states (Bavaria and Rhineland-Palatine) did not adopt it at all.8
The allocation of ethnic German immigrants across counties was not handled uniformly
by the states. Typically, the number of ethnic German immigrants allocated to the
counties was based on quotas that partly depended on the size of the resident population.
7
These counties were Wolfsburg, Salzgitter, Gifhorn, Nienburg/Weser, Cloppenburg, Emsland, and
Osnabr¨
uck. Dropping these counties from the analysis does not affect the results.
8
The amendment which introduced sanctions for leaving the assigned county was indeed effective.
For example, the counties of the Gifhorn Declaration received only few ethnic German immigrants after
the law was adapted in March 1996 (Wenzel, 1999). Glitz (2012, p. 181) furthermore notes that the
perception at the Ministry of the Interior and the Association of German Cities and Towns was that the
sanctions were effective and ensured high compliance with the initial allocation decision.
5
The allocation also depended on other factors, such as intake capacities or the size of the
county area in some states (e.g. North Rhine-Westphalia). Some states adjusted their
quotas to changes in the counties’ population size, while other states did not. Table A-1
in the Appendix indicates that ethnic Germans were indeed allocated according to these
official criteria. Counties with larger populations and counties with higher GDP per capita
(although statistically insignificant) received more ethnic Germans. Because previous
inflows reflect time-persistent county characteristics that are not captured by the other
covariates, the compound inflow in the previous three years is a predictor of the current
inflow.9
The actual county of residence of new ethnic German immigrants was largely
determined by family ties. While German authorities determined the number of ethnic
German immigrants allocated to a particular county in a given year, each new immigrant
could express her preferences to live in the county were her relatives were already living. If
the quota for the desired county was not already exhausted, the authorities tried to meet
the immigrant’s preferences. The Ministry of the Interior estimates that the presence of
family members was the decisive factor for the allocation decision in about 90% of all
cases. The presence of health and care facilities and the infrastructure for single parents
were additional factors. Most importantly, however, the skill level or the previous crime
history of ethnic German immigrants did not play any role in the allocation process (Glitz,
2012, p. 180).
Even though official data are lacking10 , there is strong evidence that the vast majority
of ethnic German immigrants stayed in their assigned county of residence during the period
of retention, that is, during the first three years after immigration. First, the vast majority
of ethnic German immigrants depended on social welfare, thus rendering departure from
the assigned county very costly. In a survey among 1,554 ethnic German immigrants,
only 11% reported that they did not receive any type of social benefit or state assistance
during the first three years after arrival (Haug and Sauer, 2007, p. 101). Furthermore,
only 36% of ethnic Germans worked at some point during the first three years after arrival
(Haug and Sauer, 2007, p. 120). Second, ethnic German immigrants were likely to comply
with the assignment because they depended on state-financed housing. After arriving in
Germany, about 80% of ethnic German immigrants lived in temporary residential homes
or admission centers (Seifert, 1996), staying on average about two years in the temporary
9
One potential issue—frequently discussed in the migration literature (Borjas, 1994)—is that residents
move to other counties in response to immigrant inflows. In our case, potential outflows of residents
would be problematic if individuals with high (low) crime propensities would move, thus leading to an
underestimation (overestimation) of the true crime effect. In line with findings that Germany has a rather
inflexible labor market (Pischke and Velling 1997), Glitz (2012) has shown that there is no systematic
outflow of residents in response to the inflow of ethnic Germans.
10
Because ethnic German immigrants cannot be identified in the registration offices of the municipalities
due to their German citizenship, the number of ethnic German immigrants residing in a given county is
unknown.
6
residential homes (Mammey, 1999). Adolescent ethnic German immigrants are also very
likely (about 51%) to live in residential establishments or in social housing, compared to
10% of native Germans (Dietz, 1999, Table 12).
Concerning socioeconomic characteristics, ethnic German immigrants have worse
outcomes than native Germans. One reason is that the share of ethnic German immigrants
of own German descent was only 47.7% in 1996 and declined even further to only 21.5%
in 2005 (Haug and Sauer, 2006, p. 417). Accordingly, the German language proficiency of
ethnic Germans is rather low. Another factor is that ethnic Germans had to assimilate in
their home countries, thereby losing German language skills. Even among ethnic German
immigrants who immigrated in the early 1990s—when the majority was of own German
descent—self-assessed German language proficiency at the end of the 1990s was not better
than the language proficiency of other immigrants who immigrated in the same years, and
was even worse than the language proficiency of second-generation immigrants in Germany
(Haug, 2005).11
Young ethnic German immigrants also tend to have low education. Administrative
school data from the largest German state, North Rhine-Westphalia, for the school year
1996/1997 show that children from ethnic Germans are highly overrepresented in the least
academic secondary school track (Hauptschule) (Dietz, 1999, Table 9). The occupational
composition of newly arriving ethnic German immigrants also indicates a rather loweducated immigrant group. Ethnic German immigrants are highly overrepresented in
low-skill occupations such as farmers, laborers, transport workers, operatives, and craft
workers (see Glitz 2012, p. 188f.).
Statistics based on the German Mircocensus 2008, the first large-scale survey that
allows identifying ethnic German immigrants unambiguously, shows that ethnic German
immigrants have much worse socioeconomic outcomes than native Germans, but similar
outcomes than other immigrants in Germany (see Table A-2 in the Appendix). Both
ethnic German immigrants and other immigrants are heavily overrepresented in the two
lowest education categories (Panel B). While ethnic German immigrants have somewhat
better employment outcomes than other immigrants, both groups have substantially worse
outcomes than native Germans, in particular much higher unemployment rates and higher
incidences of temporary employment (Panel C). Ethnic German immigrants and other
immigrants have lower incomes than native Germans (Panel D). Finally, ethnic German
immigrants and other immigrants are much more likely to receive social assistance (Hartz
IV) and unemployment benefits (ALG I) than natives.
In sum, although ethnic German immigrants are of German descent, they have
much poorer socioeconomic characteristics and worse employment outcomes than native
11
The low German language proficiency might partly be due to the fact that only few ethnic Germans
had native Germans as friends (Dietz, 1999). One reason for the lack of friendships might be the negative
attitude of native Germans toward ethnic Germans, which was sometimes more negative than toward
foreigners (Fertig and Schmidt, 2001).
7
Germans. Employment outcomes of ethnic German immigrants and other immigrant
groups in Germany, however, are rather similar.
3
Data
We have collected administrative data on reported crime at the county level (NUTS-3)
from the State Offices of Criminal Investigation (Landeskriminal¨amter) for the years 1997–
2002 and from the Federal Criminal Police Office (Bundeskriminalamt) for the years 2003–
2006. We were able to collect data on total crime and several important subcategories:
battery, drug offense, property damage, burglary, and street crime.12
Table A-3 in the Appendix reports the states used in the analysis along with county-byyear observation numbers for the different crime categories. We include all West German
states that allocated ethnic German immigrants across counties.13 The first year in our
sample is 1996 because the allocation was not binding before, and the last year is 2005
since ethnic German inflows into Germany were rather small afterwards (see Figure 1).14
Table 1 reports summary statistics of the county-by-year data. Panel A reports the
crime rates for total crime and the five subcategories. Crime rates are defined as the
number of crimes reported to the police per 100,000 inhabitants:
Crime ratet =
reported crimest × 100, 000
populationt−1
(1)
The average total crime rate in our sample is 6,883 crimes per 100,000 inhabitants.
Compared to the mean, the standard deviation of 2,661 crimes is large, which is mainly
driven by large differences between counties. After taking out permanent differences
between counties and yearly fluctuations within states by regressing the crime rate on
county and year-by-state fixed effects, the variations in crime rates decrease substantially
(see Adj. SD).
12
Total crime contains any type of crime reported to the police. Battery includes, among others, bodily
harm with fatal consequences, grievous and serious bodily harm, and bodily injury caused by negligence.
Drug offense contains all crimes associated with general violations of the narcotics law, drugs trafficking,
and illegal importation of drugs. Property damage covers all crimes that involve destroying, damaging
or making useless another person’s property. Burglary contains all crimes that include housebreaking
for committing theft. Street crime is a mixed category that includes various street-related crimes, for
example, theft of, in, and out of cars and robberies. Crime categories can be reviewed in the List of
Crimes (Straftatenkatalog) from the Federal Criminal Police Office.
13
The two states Bavaria and Rhineland-Palatinate did not allocate ethnic German immigrants across
counties. We focus on West Germany (excluding Berlin) because crime data at the county level are not
available for East German states for the years 1997–2002.
14
Hamburg starts only in 1999 due to missing information on ethnic German immigrant inflows, and
Hesse starts only in 2003 because the allocation across counties started only in 2002. Observation numbers
vary by crime category due to missing data. Furthermore, crime data for North Rhine-Westphalia were
only provided at the level of police administration areas. While in most cases police administration
areas are identical with the borders of counties, we had to aggregate 14 counties (8.3%) into 6 police
administration areas.
8
The relative importance of each crime category can be expressed by the number of
crimes in the respective category as a percentage of total crime. The largest crime
category is street crime covering 27.1% of all reported crimes. The second biggest category
is property damage with 11.0%, followed by battery (7.2%), drug offense (4.2%), and
burglary (1.9%). Taken together, these five crime categories account for 51.4% of all
reported crimes.
Annual county-level data on the inflows of ethnic German immigrants come from two
sources. Inflow data from 1996 to 2001 is taken from Glitz (2012) which originally come
from the admission centers in each state. Inflow data from 2002 to 2006 come from the
Bundesarbeitsgemeinschaft Ev. Jugendsozialarbeit e.V., Jugendmigrationsdienste which
is a church initiative that takes care of underprivileged youths. This initiative collected the
data directly from the state Ministries of the Interior. We checked the comparability of the
two different data sources by comparing the data for the overlapping years (1998–2001),
finding a very high degree of consistency.
Summary statistics of ethnic German immigrant inflows across counties are reported
in Panel B of Table 1. The average yearly inflow between 1996 and 2005 was 293 ethnic
German immigrants per county, with a minimum of 0 and a maximum of 2,393 immigrants.
There is substantial variation across counties, with a standard deviation of 270 ethnic
German immigrants.
Taking out time-invariant county differences and year-by-state
fluctuations, the adjusted standard deviation (Adj. SD) is 93 ethnic German immigrants.
The main explanatory variable of interest is the ethnic German inflow rate. The inflow
rate is defined as the number of ethnic German immigrants assigned to a particular county
in year t divided by the population of that county at the end of year t − 1:15
number of allocated ethnic German immigrantst
populationt−1
(2)
Across all counties and all years, the average inflow rate was about one ethnic German
Ethnic German inf low ratet =
immigrant per 1,000 inhabitants. Over the entire sample period (1996–2005), the average
inflow rate was 12 ethnic German immigrants per 1,000 inhabitants, that is, more than
one percent of Germany’s population. This inflow rate implies that the average county
received 3,151 ethnic German immigrants between 1996 and 2005. Because quotas were
not only based on population size, inflow rates do vary across counties. For example,
combining all years in our sample period, the inflow rate varies between 2 ethnic German
immigrants per 1,000 inhabitants and 22 ethnic German immigrants per 1,000 inhabitants.
15
Alternatively defining the inflow rate as the number of assigned ethnic Germans divided by the
county’s population plus the number of assigned ethnic Germans yields very similar results. Using the
absolute ethnic German inflow, that is, the number of assigned ethnic Germans, leads to very similar
results (not reported).
9
Summary statistics of additional county characteristics are reported in Panel C of
Table 1, including demographic information such as population, share of foreigners, and
share of young male adults aged 15 to 39. Unemployment rate and GDP per capita
reflect the regional labor market conditions. Finally, we use crime-specific clear-up rates
as proxies for the probability of being arrested (see Panel D of Table 1), with the clear-up
rate being defined as the number of crimes that are cleared times 100 and divided by the
number of reported crimes.
4
Empirical Model
To investigate the effect of immigration on crime, we estimate the following model:
ln crime rateist = α + β ethnic German inf low rateist−1 +
Xist−1 0 γ + µi + (µt × µs ) + ist
(3)
where the logarithm of the crime rate in county i in state s in year t is regressed on the
ethnic German inflow rate of the previous year. We control for county fixed effects (µi )
to eliminate any time-invariant determinants of crime across counties. Furthermore, we
include year-by-state fixed effects (µt × µs ) to net out crime trends that are common to
all counties within a state.16 Xist−1 includes lagged time-varying county characteristics
such as demographics and economic conditions.
It is a well-known problem in the crime literature that reported crimes underestimate
the true (but unobserved) number of committed crimes (e.g., Bianchi et al., 2012). This
may lead to biased estimates if the extent of underreporting is correlated with the
determinants of crime. For this reason, crime studies use the logarithm of the crime
rate and include region and year fixed effects in the regression model (Levitt, 1996; Gould
¨
et al., 2002; Oster
and Agell, 2007; Foug`ere et al., 2009; Bianchi et al., 2012). Using the
logarithm of the crime rate and including these fixed effects eliminates any measurement
error that is constant within regions over time and any measurement error that is constant
across regions in any time period. This arguably eliminates various important sources of
underreporting, such as cultural differences across regions or changing crime-awareness
over time (e.g. transported via national media).
16
Note that this model is more demanding, but also more realistic, than a model that includes only year
fixed effects. There are three reasons for including year-by-state fixed effects. First, each state decides
how to allocate ethnic German immigrants across its counties. Therefore, changes in the state-specific
allocation rule would affect all counties within the same state, but would not affect counties in other
states. Second, states, not the federal level, have the authority over police issues. For example, each
state decides about hiring new police officers and how to allocate police officers across counties. Third,
states differ in the proximity to neighboring countries which might lead to different crime developments
over time. Figure 2 presents the crime trends for six states in our sample, showing that state-specific
crime rates develop differently over time. These differing crime trends are taken into account by including
year-by-state fixed effects.
10
We cluster standard errors at the level of the included fixed effects, that is, we assume
the following error structure: ist = θi + (θt × θs ) + ηist . We allow the error term to be
correlated within counties over time (θi ) and within year-by-state cells across counties
(θt × θs ). Clustering standard errors at the year-by-state level is motivated by the fact
that states are responsible for allocating both ethnic German immigrants and police
forces across counties. We use two-way clustered standard errors to account for the two
dimensions of correlated standard errors.17
We control for several lagged county characteristics (Xist−1 ) that might affect crime
rates. Demographic characteristics include the logarithm of the population18 , the share
of foreigners, and the share of young male adults (aged 15-39) since this population group
is particularly likely to commit crimes (Freeman, 1999). We also include the logarithm
of GDP per capita and the unemployment rate to control for labor market opportunities.
Finally, we include the clear-up rate for the respective crime category as a proxy for the
expected costs of crime (see Ehrlich, 1996). Note that all county characteristics are also
lagged by one year since the inflow of ethnic German immigrants in year t − 1 might affect
these characteristics in year t, thus rendering county characteristics in year t bad controls
(Angrist and Pischke, 2009).19
Given that the allocation of ethnic German immigrants was binding for three years,
we are able to identify the short-run impact of immigration on crime. Because we exploit
annual inflow data—and because the stock of ethnic German immigrants is not available—
we cannot identify medium- to long-run effects.
By relating the crime rate in year t to the migrant inflow in the previous year (t−1), the
empirical model is set up to reflect this short-run perspective. Using the one-year lag of
the inflow ensures that we do not erroneously attribute crimes committed at the beginning
of year t to ethnic German immigrants who arrived at the end of year t. Depending on the
exact date of arrival during year t − 1, the model captures any crime committed during
the first or during the first two years after arrival. If immigrants have particularly poor
labor market outcomes in the first years after arrival, for example, due to occupational
downgrading, crime impacts might be stronger in the short run than in the long run.
The identification of a causal effect on crime depends on the exogeneity of the
allocation with respect to transitory regional crime conditions.
The exogeneity
requirement is likely satisfied because the main allocation criterion, as pointed out by
Glitz (2012, p. 193), was the proximity of family members and because labor market skills
did not play a role in the allocation process. Glitz provides evidence that the allocation
17
See Cameron et al. (2011) and Thompson (2011) for theoretical derivations of the two-way clustering
method and Acemoglu and Pischke (2003) for an application of two-way clustering. We use the Stata
command ivreg2 for computing two-way clustered standard errors.
18
Since the model includes county fixed effects, population size also captures population density, which
is one determinant of criminal behavior (Glaeser and Sacerdote, 1999).
19
Results do not change if we use contemporaneous county characteristics instead.
11
across states was exogenous with respect to individual characteristics “as suggested by the
overwhelming importance of family ties for the allocation decision.” He finds that the age
distributions of ethnic German immigrants across the West German states are very similar
(with standard deviations being much smaller than the corresponding standard deviations
of the overall population), which should be the case if immigrants were exogenously
allocated with respect to individual characteristics. Glitz argues that this finding also
indicates that the allocation within states was also exogenous since the allocation across
counties followed similar administrative processes and decision criteria as the previous
allocation to states. In sum, because both skills and age—two important determinants
of crime—were no factors in the allocation process of ethnic German immigrants, the
skill and age distributions of newly arriving ethnic German immigrants should be similar
across counties.20
One potential issue is that children of criminal fathers are much more likely of
being convicted than children with noncriminal fathers, with parents’ background
explaining the vast majority of this relationship (Hjalmarsson and Lindquist, 2012). For
intergenerational correlation in crime to invalidate our identification strategy, the family
members already living in Germany who are less criminal must have self-selected into
regions with low unemployment or low crime rates. To the extent that county-specific
labor market and crime conditions are persistent, we solve this issue by including county
fixed effects in the model. The only remaining worry is that individuals with low crime
propensities moved to regions with good conditions during our 9-year-sample period before
their family members immigrated. In addition, the intergenerational correlation in crime
between newly arriving ethnic German immigrants and their family members is likely
to be zero or small in our case since these families were typically split up a long time
ago (Glitz, 2012, p. 183), and therefore raised in very different environments (unlike the
parents and their offsprings in other studies).
5
Results
We first present results on the impact of immigration on total crime and on the various
subcategories (Section 5.1). We then provide evidence that the allocation of ethnic
German immigrants was not affected by regional crime rates and robustness checks
(Section 5.2). Finally, we assess whether crime effects depend on regional labor market
conditions, preexisting crime levels, or on the share of foreigners (Section 5.3).
20
Glitz (2012) correctly notes that this assumption may not hold if the skills between immigrants and
their relatives are correlated. However, because family members were typically separated a long time ago
and attended different education systems, the correlation in skills is likely to be small.
12
5.1
Main Results
Table 2 presents the impact of immigration on total crime. Controlling for county and
year-by-state fixed effects, we find that larger inflows of immigrants raises total crime the
following year (Column 1). Adding population size and GDP per capita–the two main
factors in determining the allocating of ethnic German immigrants across counties–does
not change the coefficient (Column 2). Given that county fixed effects are included, it is
not surprising that the coefficients on population size and GDP per capita are close to
zero. Adding unemployment rate, share of foreigners, share of young men, and clear-up
rate decreases the coefficient of interest only slightly (Column 3). Again, the coefficients
on these control variables are close to zero.
To interpret the coefficient on the inflow rate causally, one has to assume that the inflow
rate is exogenous to local crime conditions, conditional on all covariates. In contrast, the
effect of immigration on crime is not estimated consistently if there are factors that affect
both the inflow rate and the crime rate (e.g. labor market conditions that are not perfectly
captured by the included controls). The coefficient would be underestimated, for example,
if states allocated more ethnic German immigrants to counties with falling crime rates.21
Therefore, we add county-specific linear time trends to estimate the effect of immigration
on crime that results from deviations of linear time trends. Adding linear time trends
reduces the coefficient of interest only slightly, but inflates the standard error by more
than 50% (Column 4). However, controlling for region-specific time trends is problematic
because time trends likely pick up part of the treatment effects, and not just preexisting
trends (Wolfers, 2006). This is especially problematic in this setting because our sample
does not include any pre-treatment periods, that is, years without any inflow of ethnic
German immigrants. For this reason, we consider the specification in Column (3) as the
baseline model.
How large is the effect of immigration on crime? For expositional purposes, we compute
the effect size for a doubling of the average ethnic German inflow rate, that is, an increase
in the inflow by one immigrant per 1,000 inhabitants (or 293 immigrants per county).22
Multiplying the increase in the inflow rate (0.001) with the estimated coefficient yields an
increase in total crime by 0.8756%, or about 0.9%.
To compare this semi-elasticity with the crime elasticities reported in other studies,
we need the stock of ethnic German immigrants in Germany. The Microcensus 2008 is the
first large-scale administrative survey that allows identifying ethnic German immigrants
(and their offsprings) unambiguously based on their self-reported status. In the eight
states used in the analysis, the share of ethnic German immigrants (including offsprings)
in the total population is 5.4%, or 54 ethnic German immigrants per 1,000 residents.
21
This seems, however, not to be the case as inflow rates are completely unrelated to the lagged crime
rate (see Table 4).
22
Note that this increase is somewhat larger than the standard deviation of the inflow rate (0.0007).
13
Therefore, an increase in the inflow of one ethnic German immigrant per 1,000 residents
increases the stock of ethnic German immigrants by about 1.85% (=1/54). This implies
an elasticity for total crime of 0.49 (=0.9%/1.85%). Importantly, this elasticity is an
upper bound of the true elasticity since the stock of ethnic German immigrants was
somewhat lower during the analysis period (1996–2005). Assuming, for example, that
the stock of ethnic German immigrants was on average one percentage point lower (that
is, 4.4%)–consistent with the compound inflow rate of 1% during our analysis period–the
crime elasticity would be 0.40.
The crime elasticity is considerably larger than those reported in previous studies.
Bianchi et al. (2012) estimate a very low elasticity for total crime of about 0.03 for Italy,
finding a statistically significant effect only for robberies (of elasticity 1). Bell et al. (2013)
find an elasticity for total crime of about 0.16 for asylum seekers in England.
There are two potential explanations why we find stronger effects. The first
explanation concerns a crucial difference between the immigrant groups: Ethnic German
immigrants are granted German citizenship upon arrival, whereas immigrants studied
elsewhere typically remain foreigners. This is an important difference because immigrants
who possess the citizenship of the host country cannot be expelled from the country
in case of conviction. This lowers the expected conviction costs which, in turn, are
predicted to increase crime propensities (Becker, 1968; Spenkuch, 2013). In contrast,
the immigrant groups studied in other papers face the potential risk of being deported in
case of conviction.
A second potential explanation for the stronger crime impact in our study relates
to the horizon over which crime effects are measured. We identify short-run effects
on crime in the first two years after immigration. In contrast, other studies estimate
crime impacts in the medium to long run. Recent research, however, shows that newly
arriving immigrants suffer from occupational downgrading upon arrival, with labor market
outcomes improving only gradually over time (e.g.
Eckstein and Weiss (2004) and
Dustmann et al. (2013)). Because opportunity costs of committing crimes increases with
time spent in the destination country (due to having legitimate earnings opportunities),
one would expect that immigrants are more criminal in the first years after arrival.
Immigrants may commit crimes themselves (direct effect) and/or immigration
increases crime because residents respond to the inflow of immigrants (indirect effect).
The indirect effect could arise, for example, if labor market outcomes of previous residents
worsens due to the inflow of immigrants. Borjas et al. (2010), for example, show that
U.S. natives, especially black males, committed more crimes in response to increased labor
market competition with immigrants. This might also be one possible mechanism in our
case as Glitz (2012) has shown that ethnic German immigrants displace native Germans
in the labor market, with 3.1 resident workers becoming unemployed for every 10 migrants
who find a job.
14
Because ethnic Germans typically cannot be identified in official crime statistics due to
their German nationality, we cannot directly assess whether (or how much) the increase in
crime is due to a direct or indirect effect. However, a few small-scale evaluations—where
the background of suspects or prison inmates has been recorded—indicate that ethnic
Germans are (at least partly) directly responsible for the crime increase. Among
inmates in 19 juvenile prisons in West Germany in 1998, for example, ethnic Germans
are overrepresented by 100 percent relative to their population share (Pfeiffer and
Dworschak, 1999). Official crime statistics of youth (under age 21) for one large state
show that 18.6% of German suspects of violent crime have ethnic German background
(Landeskriminalamt Baden-W¨
urttemberg, 2007). For total crime, the fraction of ethnic
German suspects is 11.6%. Finally, during the 1990s, when large waves of ethnic Germans
immigrated, the share of ethnic Germans in juvenile prisons in Baden-W¨
urttemberg
increased substantially: from 0.5% in 1993 to 19.1% in 2001 (Walter, 2002). In sum,
the small-scale surveys indicate that ethnic Germans are overrepresented among suspects
and prison inmates, suggesting that the increase in crime is at least partly due to a direct
effect.23
Table 3 reports crime effects for the several subcategories. For comparison, Panel A
replicates the effect on total crime. Overall, immigration raises crime rates in several
categories. Furthermore, almost all effects are larger than the effect on total crime. This
is possible because the five subcategories cover only half of all crimes. Based on the most
restrictive specification in Column (3), immigration has the largest impact on burglary
(Panel B). An increase in the inflow of one immigrant per 1,000 inhabitants, increases
the crime rate by 5.59% (almost six times larger than for total crime). The coefficient for
property damage (Panel C) and battery (Panel D) are three times larger than the effect
on total crime. The coefficient on drug offense (Panel E) is also large, but statistically
insignificant. For street crime (Panel F), we find a very small (0.3%) and statistically
insignificant effect.
5.2
Robustness Checks
For our identification strategy to be valid, the allocation of ethnic German immigrants
must not depend on regional crime rates. If regions with low crime rates, for example,
would systematically receive more ethnic Germans the next year, we would underestimate
the true impact of immigration on crime. To test whether this is an issue, we regress the
ethnic German inflow rate in year t on the crime rate in the previous year t − 1. Following
the baseline model, we condition on county and year-by-state fixed effects. Throughout
various specifications with different sets of control variables, the ethnic German inflow
rate in the current year is completely unrelated to the crime level in the previous year
23
If the entire increase in crime was due to a direct effect, the average newly-immigrating ethnic German
would raise total crime by 0.6 crimes (evaluated at the mean of 6,883 crimes per 100,000 inhabitants).
15
(Table 4). This finding indicates that ethnic Germans have not been allocated on the
basis of the crime level in a region.
Next, we show that the baseline results are robust to alternative specifications
(Table 5). The baseline model is replicated in Column (1) for comparison. As a first
robustness check, we weight each county-year observation with the current population
size, thus giving more weight to counties with more inhabitants (Column 2). Using
population weights does not change the coefficient of interest.
Crime rates within counties are persistent because of county-specific factors that
change only slowly over time (e.g., criminal gangs living there). If these (unobserved)
factors also influence the allocation of ethnic German immigrants, this would lead to
biased estimates. To control for this potential confound directly, we add the lagged crime
rate to the baseline model. Column (3) shows that controlling for the lagged crime rate
does not change the results. This is consistent with the results in Table 4, which suggest
that the allocation of ethnic Germans does not depend on the lagged crime rate.
Instead of controlling for the lagged crime rate and including county fixed effects,
one can also exploit within-county variation in the crime rate by using a first-differenced
model:
∆ ln crime rateist =α + β ethnic German inf low rateist−1
0
+ ∆Xist
γ + (µt × µs ) + ist
(4)
The dependent variable is the change in crime rates across two consecutive years. Control
variables are also in first differences. This model is more restrictive than the model in
Column (3) because the coefficient on the lagged crime rate is restricted to be equal to
one. However, the first-differenced model yields a coefficient very similar to the baseline
model. Standard errors, however, increase since the model explains much less of the
variation in the first-differenced outcome variable (Column 4).24
Finally, we assess whether results are robust to aggregating inflows of ethnic Germans,
rather than exploiting more volatile yearly inflows. First, we aggregate the inflow rates
of the previous three years, that is, over the period the assignment to a particular county
was binding.25 This specification mitigates idiosyncratic shocks in the lagged inflow rate.
It turns out that the coefficient on the three-year ethnic German inflow rate is very similar
to the baseline estimate (Column 5).
Furthermore, we aggregate inflows by using two-year averages of all variables.
Specifically, we compute mean crime rates for 1997/1998, 1999/2000, 2001/2002,
24
A further disadvantage of the first-differenced model is that the lagged ethnic German inflow rate is,
by construction, correlated with current population and all other control variables because the covariates
are based either on current or lagged population in the denominator. Therefore, the differenced control
variables are to some extent bad controls since they are measured after the ethnic German inflow.
25
The variable is constructed by summing the total ethnic German inflow of the previous three years
((t − 1) + (t − 2) + (t − 3)) and dividing the sum by population in year t − 4. Note that using the inflows
in the previous three years means that observations of the first two sample years (1997 and 1998) have
to be dropped.
16
2003/2004, and 2005/2006. Similarly, all control variables are averaged in the prior
years (with the first observation starting in 1996/1997). This procedure should reduce
measurement error in the relationship between the ethnic German inflow rate and the
crime rate because it purges out idiosyncratic shocks of both variables. Column (6)
shows that the coefficient of interest increases slightly, suggesting that measurement error
might be only a minor issue. Because the number of observations is reduced by half,
standard errors increase substantially, rendering the coefficient (marginally) statistically
insignificant.
Another worry might be that counties with conservative political majorities are
strongly opposed to receiving ethnic German immigrants. Accordingly, conservative
counties might lobby the state government for reducing the number of ethnic Germans
assigned to them. Since each county ultimately decides how their police resources are
actually used (potentially affecting crime detection probabilities), the political majority
in a county is a potential confounding factor. As we include county fixed effects in the
analyses, only changes in political majorities over time are potentially problematic. We
address this issue by adding several distinct measures of political majorities in our main
model, e.g. by adding a dummy whether the same party has the majority in a county
and the state. The results in Table A-4 indicate that the party majority in a county does
not affect our baseline estimates.
One further issue might be that immigrants commit crimes in other counties than their
county of residence. We use spatial lags to control for these potential regional spillovers
(see Bianchi et al., 2012; Gibbons, 2004; Zenou, 2003). Spatial lags are distance-weighted
averages of crime rates in neighboring regions, thus allowing for the dependence of crime
rates between counties. More precisely, the spatial lag is the log of the sum of the total
crime rate in all other counties, weighted by the inverse of the Euclidean geographic
distance between county centroids in kilometers (distance km) and the travel time distance
between county capitals in car minutes (travel distance min), respectively. Each regression
controls for the full set of control variables of the baseline specification. Including spatial
lags does not change the results (see Table A-5). The coefficients on both versions of
spatial lags are small and statistically insignificant.26 This is consistent with official
statistics on criminal youth in one large German state (Baden-W¨
urttemberg), which
suggest that ethnic German immigrants are more likely to commit crime in their county
of residence than native Germans or foreigners (Landeskriminalamt Baden-W¨
urttemberg,
27
2007, p. 28).
26
Of course, we cannot use crime rates of neighboring counties that did not provide crime rates at the
county level. Thus, for computing spatial lags, we are not using counties in Bavaria, Rhineland-Palatinate,
Berlin or East Germany. For these regions, we have to impose the assumption that there are no spillover
effects. However, we observe county-level crime rates of neighboring counties for the vast majority of
counties in our sample.
27
In general, crimes in Germany are committed to a large extent in the county of residence. For example,
the Federal Criminal Police Office (Bundeskriminalamt) reports that about 75 percent of all crimes
17
5.3
Effect Heterogeneity
So far, we have estimated at the average impact of immigration on crime across all
counties. Given that the allocation of ethnic German immigrants across counties was likely
homogenous with respect to individual characteristics (see Section 2), heterogeneous crime
effects across regions can therefore be attributed to differences across counties (rather
than to differences between ethnic German immigrants). We can therefore exploit this
allocation policy to assess whether the impact of immigration on crime depends on the
characteristics of the region.
We first investigate the importance of local labor market conditions. According to the
economic model of crime, we expect that the effect of immigration on crime is smaller
in regions with better labor market conditions because better legitimate employment
opportunities increase the opportunity costs of crime.
To start with, we investigate whether the effect of immigration on crime differs between
regions with different unemployment levels. To do so, we characterize counties by quartiles
in the county-level unemployment rate distribution separately for each year. Counties with
an unemployment rate below the 25th percentile in year t − 1, for example, are indicated
by I(unemployment ratet−1 ≤ 25th percentile).
Panel A in Table 6 indicates that counties with low levels of unemployment (below 25th
percentile) experience no crime increase with increasing immigrant inflows (Column 1).
In contrast, the crime impact of immigrants in counties with high unemployment (above
75th percentile) is 1.5% (sum of the coefficients 6.1+9.3). This effect is almost twice as
large as the average effect on crime (0.9%). These results indicate that immigration does
not increase crime in regions with low unemployment levels, but only increases crime in
regions with higher unemployment rates.
The crime-enhancing effect of bad labor market conditions is consistent with the
predictions of the economic model of crime (see Becker, 1968; Ehrlich, 1973): crime
impacts are stronger in regions where legitimate earnings opportunities are scarce. The
results are also in line with empirical studies showing that improvements in regional labor
market conditions reduce crime rates. For example, Gould et al. (2002) and Raphael and
Winter-Ebmer (2001) provide evidence for the U.S. that lower unemployment rates in
local labor markets decrease the incidence of crimes, especially among less educated men.
¨
Oster
and Agell (2007) find for Sweden that a decrease of municipality-level unemployment
decreases crime rates. Foug`ere et al. (2009) provide region-level evidence for France that
increases in youth unemployment increases crime.
In further analysis (available upon request), we find no evidence that crime effects
differ across counties with different levels of GDP per capita. One interpretation is that
GDP per capita is a poor proxy for employment opportunities. Consistent with this
reported in 2006 were committed in the county of the residence of the suspects (Bundeskriminalamt,
2006, Table 21).
18
interpretation, the cross-county correlation between GDP per capita and unemployment
is very low in our sample, likely because unemployment is much more volatile than GDP
per capita.
To assess whether the crime impact is stronger in regions with preexisting high crime
levels, we interact the immigration inflow with an indicator for whether the county has low
or high crime levels. Column (1) in Panel B suggests that the crime impact of immigration
is rather small in counties with low crime levels. In contrast, crime effects are strong in
regions with very high crime levels (Column 3).
The stronger crime impact in regions with high crime levels is consistent with
social interactions. Existing research indicates that social interactions is an important
explanation for criminal behavior. Damm and Dustmann (2014), for example, find that
early exposure to neighborhood crime, as measured by the share of young people convicted
for crimes, increases convictions of male adolescents. Similarly, juvenile offenders who
serve time in the same correctional facility affect each others criminal activities upon
release (Bayer et al., 2009). Glaeser et al. (1996) provide evidence that the large cross-city
variance in crime rates in the U.S. can likely not be explained by differences in costs and
benefits of crime, but that social interactions have to be considerable, especially for petty
and moderate crimes. Furthermore, Zenou (2003) shows that individuals are more likely
to commit crimes if their peers are criminal. Given this evidence, interactions between
newly arriving ethnic Germans and criminal residents might explain why crime effects are
stronger in regions with higher preexisting crime levels.
Finally, we investigate whether crime effects differ across regions with different shares
of foreigners. This analysis is motivated by the observation that foreigners are more
likely to commit crimes than natives (Bundeskriminalamt, 2006).28 The pattern is similar
to that of preexisting crime levels (Panel C): Crime effects are much larger in regions
with larger shares of foreigners. Given that foreigners are on average more criminal than
natives, social interactions might also drive this pattern.
6
Conclusion
The criminal behavior of immigrants is a huge concern among residents in many
industrialized countries. Despite this fact, little is known about the impact of immigration
on crime. This paper contributes to this literature, with a focus on the importance of
regional labor market conditions.
We investigate the crime effect of ethnic German immigrants who were allocated
across regions by German authorities upon arrival. Because this allocation was based
on the presence of family members, and because noncompliance was strongly sanctioned,
28
Although crime rates of natives are not directly comparable to those of foreigners
(Bundeskriminalamt, 2006, p. 76), foreigners are highly overrepresented among suspects.
19
the possibility of self-selection into regions was severely limited. Therefore, inflows into
regions were likely not driven by local crime or local labor market conditions. The
allocation policy thus provides a unique quasi-experimental setting for studying the effects
of immigration on crime.
We find that the inflow of ethnic German immigrants strongly increased crime rates.
Importantly, we find that the crime impact depends heavily on the labor market condition
in a region: In regions with low unemployment, immigration does not increase crime. In
contrast, in regions with high unemployment, immigration raises crime rates substantially.
The crime-enhancing effect of bad labor market conditions is consistent both with the
economic theory of crime and with empirical studies finding that improvements in local
labor market conditions reduce crime. Furthermore, the impact of immigration on crime
seems particularly strong in regions with high preexisting crime levels and in regions with
large foreigner shares—findings that point toward social interactions.
Existing studies tend to find zero or small effects of immigration on crime. In contrast,
this study finds large effects, suggesting that characteristics of immigrant groups are
crucial for crime effects. In particular, immigrants in this study were not threatened with
deportation in case of conviction. Furthermore, the findings suggest that poorly educated
migrants with deficiencies in the host country language are particularly vulnerable in
economically disadvantaged regions. Our study therefore cautions about comparing crime
effects across countries since both labor market conditions and immigrant characteristics
might differ substantially. Our results furthermore indicate that a successful integration
of immigrants into the labor market immediately after arrival seems crucial for crime
prevention.
20
References
Acemoglu, D. and Pischke, J.-S. (2003). Minimum Wages and On-the-job Training.
Research in Labor Economics, 22:159–202.
Angrist, J. D. and Pischke, J.-S. (2009). Mostly Harmless Econometrics: An Empiricist’s
Companion. Princeton University Press.
Bade, K. J. (1990). Aussiedler - R¨
uckwanderer u
¨ber Generationen hinweg. In Neue
Heimat im Westen. Vertriebene. Fl¨
uchtlinge. Aussiedler, pages 128–149. Westf¨alischer
Heimatbund, M¨
unster.
Bade, K. J. and Oltmer, J. (2003).
Aussiedler: deutsche Einwanderer aus
Osteuropa.
In Einf¨
uhrung: Aussiedlerzuwanderung und Aussiedlerintegration.
Historische Entwicklung und aktuelle Probleme, pages 9–51. IMIS-Schriften 8,
G¨ottingen.
Bayer, P., Hjalmarsson, R., and Pozen, D. (2009). Building Criminal Capital behind Bars:
Peer Effects in Juvenile Corrections. Quarterly Journal of Economics, 124(1):105–147.
Becker, G. (1968). Crime and Punishment: An Economic Approach. Journal of Political
Economy, 76(2):169–217.
Bell, B., Fasani, F., and Machin, S. (2013). Crime and Immigration: Evidence from Large
Immigrant Waves. Review of Economics and Statistics, 95(4):1278–1290.
Bianchi, M., Buonanno, P., and Pinotti, P. (2012). Do Immigrants Cause Crime? Journal
of the European Economic Association, 10(6):1318–1347.
Borjas, G. (1994). The Economics of Immigration. Journal of Economic Literature,
32(4):1667–1717.
Borjas, G. J. (1999). The Economic Analysis of Immigration. In Handbook of Labor
Economics, volume 3A, edited by Orley C. Ashenfelter and David Card, chapter 28,
pages 1697–1760. Elsevier, North-Holland, Amsterdam.
Bundeskriminalamt (2006). Polizeiliche Kriminalstatistik 2006.
Butcher, K. and Piehl, A. M. (1998). Recent Immigrants: Unexpected Implications for
Crime and Incarceration. Industrial and Labor Relations Review, 51(4):654–679.
Butcher, K. and Piehl, A. M. (2007). Why are Immigrants’ Incarceration Rates so Low?
Evidence on Selective Immigration, Deterrence, and Deportation. NBER Working
Paper No. 13229.
Camarota, S. and Vaughan, J. (2009). Immigration and Crime: Assessing a Conflicted
Issue. Washington, DC: Center for Immigration Studies.
Cameron, A. C., Gelbach, J. B., and Miller, D. L. (2011). Robust Inference with Multiway
Clustering. Journal of Business & Economic Statistics, 29(2):238–249.
Card, D. (2001). Immigrant Inflows, Native Outflows, and the Local Labor Market
Impacts of Higher Immigration. Journal of Labor Economics, 19(1):22–64.
Card, D. (2009). How Immigration Affects U.S. Cities. In Making Cities Work: Prospects
and Policies for Urban America, edited by Robert P. Inman, chapter 6, pages 158–200.
Princeton University Press, Princeton.
21
Card, D., Dustmann, C., and Preston, I. (2012). Immigration, Wages, and Compositional
Amenities. Journal of the European Economic Association, 10(1):78–119.
Chalfin, A. (2013). What is the Contribution of Mexican Immigration to U.S. Crime
Rates? Evidence from Rainfall Shocks in Mexico. American Law and Economics
Review, 16(1):220–268.
Damm, A. P. (2009). Ethnic Enclaves and Immigrant Labor Market Outcomes: QuasiExperimental Evidence. Journal of Labor Economics, 27(2):281–314.
Damm, A. P. and Dustmann, C. (2014). Does Growing Up in a High Crime Neighborhood
Affect Youth Criminal Behavior? American Economic Review, 104(6):1806–1832.
Dietz, B. (1999). Jugendliche Aussiedler in Deutschland: Risiken und Chancen der
Integration. In Bade, K. J. and Oltmer, J., editors, Aussiedler: Deutsche Einwanderer
aus Osteuropa, pages 153–176. IMIS-Schriften 8, G¨ottingen.
Dustmann, C., Frattini, T., and Preston, I. (2013). The Effect of Immigration along the
Distribution of Wages. Review of Economic Studies, 80(1):145–173.
Eckstein, Z. and Weiss, Y. (2004). On the Wage Growth of Immigrants: Israel, 1990-2000.
Journal of the European Economic Association, 2(4):665–695.
Edin, P.-A., Fredriksson, P., and ˚
Aslund, O. (2003). Ethnic Enclaves and the Economic
Success of Immigrants – Evidence from a Natural Experiment. Quarterly Journal of
Economics, 118(1):329–357.
Ehrlich, I. (1973). Participation in Illegitimate Activities: A Theoretical and Empirical
Investigation. Journal of Political Economy, 81(3):521–565.
Ehrlich, I. (1996). Crime, Punishment, and the Market for Offenses. Journal of Economic
Perspectives, 10(1):43–67.
Fertig, M. and Schmidt, C. M. (2001). First- and Second-Generation Migrants in Germany
- What Do We Know and What Do People Think. IZA Discussion Paper No. 286.
Fitzgerald, J., Curtis, K. A., and Corliss, C. L. (2012). Anxious Publics: Worries about
Crime and Immigration. Comparative Political Studies, 45(4):477–506.
Foug`ere, D., Kramarz, F., and Pouget, J. (2009). Youth Unemployment and Crime in
France. Journal of the European Economic Association, 7(5):909–938.
Freeman, R. (1999). The Economics of Crime. In Handbook of Labor Economics, volume
3C, edited by Orley C. Ashenfelter and David Card, chapter 52, pages 3529–3571.
Elsevier, North-Holland, Amsterdam.
Friedberg, R. and Hunt, J. (1995). The Impact of Immigrants on Host Country Wages,
Employment and Growth. Journal of Economic Perspectives, 9(2):23–44.
Gibbons, S. (2004). The Costs of Urban Property Crime.
114(499):F441–F463.
Economic Journal,
Glaeser, E., Sacerdote, B., and Scheinkman, J. (1996). Crime and Social Interactions.
Quarterly Journal of Economics, 111(2):507–548.
Glaeser, E. L. and Sacerdote, B. (1999). Why Is there More Crime in Cities? Journal of
Political Economy, 107(S6):S225–S258.
22
Glitz, A. (2012). The Labor Market Impact of Immigration: A Quasi-Experiement
Exploiting Immigrant Location Rules in Germany. Journal of Labor Economics,
30(1):175–213.
Gould, E. D., Weinberg, B. A., and Mustard, D. B. (2002). Crime Rates and Local
Labor Market Opportunities in the United States: 1979-1997. Review of Economics
and Statistics, 84(1):45–61.
Haug, S. (2005). Zum Verlauf des Zweitspracherwerbs im Migrationskontext. Zeitschrift
f¨
ur Erziehungswissenschaft, 8(2):263–284.
Haug, S. and Sauer, L. (2006). Zuwanderung und r¨aumliche Verteilung von Aussiedlern
und Sp¨ataussiedlern in Deutschland. Zeitschrift f¨
ur Bev¨olkerungswissenschaft, 31(34):413–442.
Haug, S. and Sauer, L. (2007). Zuwanderung und Integration von (Sp¨at-) Aussiedlern
- Ermittlung und Bewertung der Auswirkungen des Wohnortzuweisungsgesetzes.
Forschungsbericht 3, Bundesamt f¨
ur Migration und Fl¨
uchtlinge, N¨
urnberg.
Hjalmarsson, R. and Lindquist, M. J. (2012). Like Godfather, Like Son: Exploring the
Intergenerational Nature of Crime. Journal of Human Resources, 47(2):550–582.
Landeskriminalamt Baden-W¨
urttemberg (2007).
Jugendkriminalit¨at und
Jugendgef¨ahrdung in Baden W¨
urttemberg. Jahresbericht 2006, Landeskriminalamt
Baden-W¨
urttemberg, Stuttgart.
Levitt, S. (1996). The Effect of Prison Population Size on Crime Rates: Evidence from
Prison Overcrowding Litigation. Quarterly Journal of Economics, 111(2):319–351.
Mammey, U. (1999). Segregation, regionale Mobilitat und soziale Integration von
Aussiedlern. In Bade, K. J. and Oltmer, J., editors, Aussiedler: deutsche Einwanderer
aus Osteuropa, pages 107–126. IMIS-Schriften 8, G¨ottingen.
Nieders¨achsische Landeszentrale f¨
ur Politische Bildung (2002).
Hier geblieben.
Zuwanderung und Integration in Niedersachen 1945 bis heute. http://www.politischebildung.de/niedersachsen/hiergeblieben zuwanderung.pdf. Date Accessed: 12.07.2006.
¨
Oster,
A. and Agell, J. (2007). Crime and Unemployment in Turbulent Times. Journal
of the European Economic Association, 5(4):752–775.
Peters, W. (2003). Aussiedlerzuzug: Entwicklung und Perspektiven. Zeitschrift f¨
ur
Ausl¨anderrecht und Ausl¨anderpolitik, 23(5/6):193–197.
Pfeiffer, C. and Dworschak, B. (1999). Die ethnische Vielfalt in den Jugendstrafvollzugsanstalten. Ergebnisse einer Umfrage aus dem Sommer 1998. DVJJ-Journal,
2:184–188.
Raphael, S. and Winter-Ebmer, R. (2001). Identifying the Effect of Unemployment on
Crime. Journal of Law and Economics, 44(1):259–283.
Seifert, W. (1996). Neue Zuwanderergruppen auf dem westdeutschen Arbeitsmarkt: eine
Analyse der Arbeitsmarktchancen von Aussiedlern, ausl¨andischen Zuwanderern und
¨
ostdeutschen Ubersiedlern.
Soziale Welt, 2:180–201.
Spenkuch, J. L. (2013). Understanding the Impact of Immigration on Crime. American
Law and Economics Review, 16(1):177–219.
23
Swiaczny, F. and Mammey, U. (2001). Aussiedler. In Gans, P. and Kemper, F.-J.,
editors, Nationalatlas Bundesrepublik Deutschland, Band 4: Bev¨olkerung, pages 132–
135. Spektrum, Leipzig/Heidelberg.
Thompson, S. B. (2011). Simple Formulas for Standard Errors that Cluster by Both Firm
and Time. Journal of Financial Economics, 99(1):1–10.
Walter, J. (2002). Jugendvollzug in der Krise? DVJJ- Journal, 5:127–143.
Wenzel, H.-J. (1999). Aussiedlerzuwanderung als Strukturproblem in l¨andlichen R¨aumen.
In Bade, K. J. and Oltmer, J., editors, Aussiedler: deutsche Einwanderer aus Osteuropa,
pages 265–281. IMIS-Schriften Band 8, G¨ottingen.
Wolfers, J. (2006). Did Unilateral Divorce Raise Divorce Rates? A Reconciliation and
New Results. American Economic Review, 96(5):1802–1820.
Zenou, Y. (2003). The Spatial Aspects of Crime. Journal of the European Economic
Association, 1(2-3):459–467.
24
Figures and Tables
Figure 1: Annual Inflow of Ethnic German Immigrants to Germany
Ethnic German inflow
400,000
300,000
200,000
100,000
0
1951 1956 1961 1966 1971 1976 1981 1986 1991 1996 2001 2006
Year
Total
USSR
Poland
Romania
Notes: The figure shows the annual inflow of ethnic German immigrants to Germany by source country.
The gray-shaded area between 1996 and 2005 highlights the period used in the analysis. Source: Federal
Office of Administration (Bundesverwaltungsamt).
Figure 2: Crime Trends by State
Total Crime
Battery
8,000
7,000
6,000
400
700
Crime Rate
Crime Rate
Crime Rate
9,000
600
500
400
300
5,000
1997
2000 2003
Year
2006
350
300
250
200
1997
Property Damage
2000
2003
Year
2006
1997
Burglary
3,000
1,000
250
2,500
600
Crime Rate
300
800
200
150
100
400
50
1997
2000 2003
Year
2006
Lower Saxony
Hesse
2000
2003
Year
2006
Street Crime
1,200
Crime Rate
Crime Rate
Drug Offense
800
2,000
1,500
1,000
1997
2000
2003
Year
NRW
Saarland
2006
1997
2000 2003
Year
2006
Baden-Württ.
Schleswig-Hol.
Notes: The figure shows crime rates per 100,000 residents by crime category and state. Battery includes
only grievous and serious bodily harm and aggravated battery. The city states Hamburg and Bremen
are excluded from the figure for expositional reasons because both states have much higher crime levels.
Source: Federal Criminal Office: Police Criminal Statistics (Polizeiliche Kriminalstatistik ), 1997–2006.
Table 1: Summary Statistics
Variable
Mean
Min
Max
SD
Adj. SD
Obs
2,661
222
176
282
83
967
516
46
65
79
26
159
1,687
1,358
1,687
1,358
1,365
1,358
270
0.0007
93
0.0004
1,687
1,687
203,034
4.16
1.33
3.12
8,724
3,274
0.29
0.17
0.58
1,076
1,687
1,687
1,687
1,687
1,687
6.9
2.9
3.3
6
15.2
5
2.4
1.3
2.0
4.1
7.5
2.8
1,687
1,358
1,687
1,358
1,365
1,358
Panel A: Crime Rates
Total crime
Battery
Drug
Property damage
Burglary
Street crime
6,883
493
291
760
133
1,862
620
58
35
254
15
466
18,569
1,727
1,730
2,375
593
5,404
Panel B: Ethnic German Immigrants
Ethnic German inflow
Ethnic German inflow rate
293
0.0011
0
0
2,393
0.0049
Panel C: County Characteristics
Population
Foreign population share
Young male population share
Unemployment rate
GDP per capita
262,615
8.94
16.76
10.1
25,376
50,878
2.67
13.09
4
12,634
1,754,182
26.28
21.41
25.2
77,318
Panel D: Clear-up Rates
Total crime
Battery
Drug
Property damage
Burglary
Street crime
53.4
90.8
96
26.2
28.2
16.8
34.8
80.4
50.9
12.5
2
6.6
75.9
99.7
104.2
70.5
91.3
43.8
Notes: Crime rates in Panel A refer to years 1997–2006. All variables in Panels B–D refer
to years 1996–2005. Crime rate is defined as the number of crimes reported to the police in
year t per 100,000 inhabitants in year t − 1. Ethnic German inflow rate is defined as the
number of allocated ethnic German immigrants in year t divided by the total population in year
t − 1. Summary statistics in Panels B–D are computed for counties with data on total crime
(see Table A-3). Young male population share is the share of 15–39 year old adults in the total
population. Young male population share, foreign population share, and unemployment rate are
reported in percent. The adjusted standard deviation (Adj. SD) is the standard deviation of the
residuals obtained in a regression of the indicated variable on county and year-by-state fixed effects.
Data sources: Panel A+D: State Offices of Criminal Investigation (Landeskriminal¨
amter) for the years
1997–2002 and Federal Criminal Police Office (Bundeskriminalamt) for the years 2003–2006. Panel B:
Inflow data for 1996–2001 is taken from Glitz (2012); inflow data for 2002–2006 come from the
Bundesarbeitsgemeinschaft Evangelische Jugendsozialarbeit e.V., Jugendmigrationsdienste. Panel C:
Population data and GDP per capita come from the Federal Statistical Office (Regionalstatistik ) and
unemployment rates from the Federal Employment Agency (Bundesagentur f¨
ur Arbeit).
Table 2: Effect of Ethnic German Immigrants on Total Crime
(1)
(2)
(3)
(4)
Dependent variable: log total crime ratet
Ethnic German inflow ratet−1
9.303**
(4.591)
9.455**
(4.599)
0.045
(0.282)
0.016
(0.061)
8.756**
(4.196)
–0.031
(0.306)
–0.010
(0.061)
–0.006
(0.005)
0.003
(0.004)
–0.021
(0.020)
–0.002
(0.001)
7.324
(6.422)
–0.086
(0.433)
0.102
(0.086)
–0.013
(0.010)
0.002
(0.003)
–0.045
(0.052)
–0.002*
(0.001)
x
x
x
x
x
x
x
x
x
0.916
1,687
0.916
1,687
0.916
1,687
0.928
1,687
Log populationt−1
Log GDP per capitat−1
Unemployment ratet−1
Foreign population sharet−1
Young male population sharet−1
Clear-up ratet−1
County fixed effects
Year-by-state fixed effects
County-specific linear time trends
R-squared
Observations
Notes: Ethnic German inflow ratet−1 is defined as the number of allocated ethnic German immigrants
in year t − 1 divided by the population in year t − 2. Robust standard errors in parentheses are two-way
clustered at the county and year-by-state level. Significance levels: *** p<0.01, ** p<0.05, * p<0.1.
Table 3: Effect of Ethnic German Immigrants on Crime Subcategories
(1)
(2)
(3)
Panel A: log total crime ratet
Ethnic German inflow ratet−1
R-squared
Observations
9.303**
(4.591)
9.455**
(4.599)
8.756**
(4.196)
0.916
1,687
0.916
1,687
0.916
1,687
Panel B: log burglary crime ratet
Ethnic German inflow ratet−1
R-squared
Observations
42.146*
(23.259)
53.779***
(20.477)
55.852***
(20.635)
0.895
1,365
0.898
1,365
0.898
1,365
Panel C: log property damage crime ratet
Ethnic German inflow ratet−1
R-squared
Observations
23.648**
(9.935)
27.553***
(9.755)
27.968***
(9.217)
0.926
1,358
0.926
1,358
0.927
1,358
Panel D: log battery crime ratet
Ethnic German inflow ratet−1
R-squared
Observations
20.228
(13.036)
24.656*
(12.715)
26.786**
(11.766)
0.945
1,358
0.946
1,358
0.946
1,358
Panel E: log drug offense crime ratet
Ethnic German inflow ratet−1
R-squared
Observations
9.355
(20.311)
16.683
(20.109)
19.807
(20.215)
0.831
1,687
0.832
1,687
0.833
1,687
Panel F: log street crime ratet
Ethnic German inflow ratet−1
R-squared
Observations
Control variables–Set 1
Control variables–Set 2
County fixed effects
Year-by-state fixed effects
5.213
(7.950)
5.362
(7.990)
3.303
(7.934)
0.978
1,358
0.978
1,358
0.979
1,358
x
x
x
x
x
x
x
x
x
Notes: Ethnic German inflow ratet−1 is defined as the number of allocated ethnic German immigrants
in year t − 1 divided by the population in year t − 2. Control variables–Set 1 contains log populationt−1
and log GDP per capitat−1 . Control variables–Set 2 contains unemployment ratet−1 , foreign population
sharet−1 , young male population sharet−1 , and crime-specific clear-up ratest−1 . Robust standard errors
in parentheses are two-way clustered at the county and year-by-state level. Significance levels: ***
p<0.01, ** p<0.05, * p<0.1.
Table 4: Testing for the Effect of Lagged Crime Rates on Current Ethnic
German Inflow Rates
(1)
(2)
(3)
(4)
Dependent variable: ethnic German inflow ratet
Log total crime ratet−1
0.00005
(0.00005)
0.00006
(0.00005)
–0.0048***
(0.0017)
0.0012***
(0.0004)
0.00005
(0.00005)
–0.0045***
(0.0017)
0.0007**
(0.0003)
–0.0001***
(0.00003)
0.0004
(0.0108)
0.00006
(0.00005)
–0.0041***
(0.0015)
0.0008***
(0.0003)
–0.0001***
(0.00003)
–0.0005
(0.0099)
0.0177
(0.0265)
x
x
x
x
x
x
x
x
0.708
1,660
0.721
1,660
0.727
1,660
0.729
1,660
Log populationt−1
Log GDP per capitat−1
Unemployment ratet−1
Foreign population sharet−1 × 10−3
Ethnic German inflow rate
of previous 3 yearst−1
County fixed effects
Year-by-state fixed effects
R-squared
Observations
Notes: Ethnic German inflow rate of previous 3 yearst−1 = ethnic German inflow ratet−1 + ethnic
German inflow ratet−2 + ethnic German inflow ratet−3 . Robust standard errors in parentheses are
two-way clustered at the county and year-by-state level. Significance levels: *** p<0.01, ** p<0.05, *
p<0.1.
Table 5: Alternative Model Specifications
(1)
(2)
(3)
(4)
(5)
(6)
Baseline
Weighted
Lag
First-diff
3-year sum
2-year average
∆ log total
crime ratet
log total crime
ratet
Average log
total crime ratet
Dependent variables:
Ethnic German inflow ratet−1
log total crime ratet
8.756**
(4.196)
8.518**
(4.080)
Log total crime ratet−1
8.543**
(4.347)
–0.020
(0.093)
8.202
(5.343)
Three-year ethnic German inflow ratet−1
8.638*
(4.740)
Two-year avg. ethnic German inflow ratet−1
Control variables
County fixed effects
Year-by-state fixed effects
R-squared
Observations
9.395
(6.454)
x
x
x
x
x
x
x
x
x
0.916
1,687
0.922
1,687
0.916
1,687
x
x
x
x
x
x
x
x
0.103
1,687
0.907
1,369
0.958
844
Notes: The dependent variable in Columns (1)–(3) and (5) is the log total crime ratet . In Column (2), county observations are weighted with current population.
In Column (4), the dependent variable is in first differences: ∆ log total crime ratet = log total crime ratet - log total crime ratet−1 . All control variables are
also in first differences. In Column (6), crime rates are averaged over two consecutive years (1997/1998, 1999/2000, 2001/2002, 2003/2004, and 2005/2006).
Similarly, all control variables are averaged in the prior years (with the first observation starting in 1996/1997). Ethnic German inflow ratet−1 is defined as the
number of allocated ethnic German immigrants in year t − 1 divided by the population in year t − 2. Three-year ethnic German inflow ratet−1 equals ethnic
German inflow in year (t − 1) + (t − 2) + (t − 3) over population in year t − 4. Two-year average ethnic German inflow ratet−1 equals the simple average of the
ethnic German inflow rate of two consecutive years, starting with the first observation in 1996/1997. Control variables include log populationt−1 , log GDP per
capitat−1 , unemployment ratet−1 , foreign population sharet−1 , young male population sharet−1 , and clear-up ratet−1 . Robust standard errors in parentheses
are two-way clustered at the county and year-by-state level. Significance levels: *** p<0.01, ** p<0.05, * p<0.1.
Table 6: Effect Heterogeneity
(1)
(2)
(3)
Dependent variable: log total crime ratet
Panel A: Interaction with Unemployment Level
Ethnic German inflow ratet−1
Inflow rate × I(unemployment ratet−1 ≤ 25th percentile)
11.369**
(4.479)
–12.393*
(6.370)
3.529
(5.838)
Inflow rate × I(unemployment ratet−1 ≥ 50th percentile)
9.243*
(5.119)
Inflow rate × I(unemployment ratet−1 ≥ 75th percentile)
Effect for counties in indicator function
6.110
(4.911)
9.303
(7.056)
-1.024
(7.109)
12.772***
(4.573)
15.413**
(6.515)
Panel B: Interaction with Preexisting Crime Level
Ethnic German inflow ratet−1
Inflow rate × I(log total crime ratet−1 ≤ 25th percentile)
9.868**
(4.346)
–4.637
(8.896)
Inflow rate × I(log total crime ratet−1 ≥ 50th percentile)
7.238
(5.003)
3.392
(4.824)
Inflow rate × I(log total crime ratet−1 ≥ 75th percentile)
Effect for counties in indicator function
6.513
(4.432)
16.410**
(8.076)
5.231
(8.604)
10.629**
(4.742)
22.922***
(8.136)
9.502*
(4.914)
7.504*
(4.227)
Panel C: Interaction with Share of Foreigners
Ethnic German inflow ratet−1
Inflow rate × I(foreigner sharet−1 ≤ 25th percentile)
10.854**
(4.491)
–7.256
(10.286)
Inflow rate × I(foreigner sharet−1 ≥ 50th percentile)
–2.491
(8.998)
Inflow rate × I(foreigner sharet−1 ≥ 75th percentile)
Effect for counties in indicator function
Control variables
County fixed effects
Year-by-state fixed effects
14.033
(12.152)
3.598
(9.236)
7.010
(7.692)
21.537*
(12.163)
x
x
x
x
x
x
x
x
x
Notes: Ethnic German inflow ratet−1 is defined as the number of allocated ethnic German immigrants
in year t − 1 divided by the population in year t − 2. Inflow rate denotes ethnic German inflow ratet−1 .
I(·) denotes an indicator variable. Control variables: log GDP per capitat−1 , unemployment ratet−1 ,
log populationt−1 , foreign population sharet−1 , young male population sharet−1 , and clear-up ratet−1 .
R-squared (=0.916) and number of observations (1,687) are the same in each model. Robust standard
errors in parentheses are two-way clustered at the county and year-by-state level. Significance levels:
*** p<0.01, ** p<0.05, * p<0.1.
Table A-1: Determinants of Ethnic German Immigrant Inflow into Counties
(1)
(2)
(3)
(4)
(5)
(6)
(7)
0.22***
(0.01)
112.61***
(28.95)
11.38
(18.06)
0.36
(1.60)
–1.57
(1.33)
0.17***
(0.02)
Dependent variable: ethnic German inflowt
Log populationt−1
306.33***
(32.54)
Log GDP per capitat−1
102.81
(69.47)
Unemployment ratet−1
–6.28
(6.51)
Foreign population sharet−1
13.03**
(5.18)
329.12***
(35.29)
57.99
(43.47)
0.64
(3.42)
–9.47***
(3.31)
Ethnic German inflow of previous 3 yearst−1
Year fixed effects
R-squared
Observations
x
x
x
x
x
x
x
0.618
1,660
0.147
1,660
0.140
1,660
0.177
1,660
0.630
1,660
0.767
1,660
0.795
1,660
Notes: Ethnic German inflowt is the number of ethnic German immigrants allocated to a given county in year t. Robust standard errors in parentheses are
two-way clustered at the county and year-by-state level. Significance levels: *** p<0.01, ** p<0.05, * p<0.1.
Table A-2: Characteristics of Ethnic German Immigrants,
Other Immigrants, and Native Germans
Ethnic Germans –
Variable
Ethnic
Germans
Other
Immigrants
Native
Germans
Other
Immigrants
Native
Germans
1.9***
[0.352]
0.03***
[0.008]
1.1***
[0.076]
0.857***
[0.0002]
-9.81***
[0.177]
0.00
[0.001]
n/a
-0.069***
[0.009]
0.048***
[0.01]
0.148***
[0.011]
-0.127***
[0.013]
0.087***
[0.003]
0.126***
[0.006]
-0.089***
[0.009]
-0.128***
[0.010]
-0.014
[0.016]
0.041**
[0.017]
0.089***
[0.014]
-1.4***
[0.274]
-0.006
[0.016]
0.080***
[0.007]
-0.067***
[0.009]
-0.018***
[0.006]
-2.1***
[0.247]
0.106***
[0.008]
-0.060***
[0.018]
0.055***
[0.019]
0.054***
[0.015]
0.001
[0.013]
-0.029***
[0.009]
-0.022***
[0.005]
0.071***
[0.002]
0.206***
[0.013]
0.029**
[0.011]
-0.094***
[0.01]
-0.144***
[0.01]
-0.070***
[0.006]
-0.016
[0.018]
-0.007
[0.007]
0.127***
[0.007]
0.019***
[0.003]
Panel A: Demographics
Age
Male
Years since migration
German citizenship
35.0
(22.3)
0.49
33.1
(17.5)
0.46
44.8
(11.3)
0.49
8.5
(4.6)
1
7.4
(3.9)
0.143
n/a
1
0
Panel B: Education
ISCED 1
0.107
0.176
0.017
ISCED 2
0.298
0.250
0.172
ISCED 3
0.439
0.291
0.528
ISCED 4-6
0.156
0.283
0.284
Panel C: Employment
Unemployment rate
0.127
0.141
0.047
Employment ratio
0.820
0.779
0.887
Full-time working
0.933
0.844
0.951
Usual hours worked
40.7
(7.5)
0.165
42.1
(11.6)
0.171
42.8
(11.5)
0.059
Temporary
Panel D: Net income
< 900
0.125
0.185
0.054
900-1,500
0.428
0.373
0.222
1,500-2,000
0.286
0.232
0.257
2,000-2,600
0.122
0.121
0.216
2,600-4,000
0.031
0.060
0.175
> 4,000
0.007
0.029
0.077
Panel E: Social Assistance
Hartz IV
0.178
0.194
0.051
ALG I
0.035
0.042
0.016
Notes: The samples of ethnic Germans and other immigrants include only individuals who immigrated to Germany
between 1996 and 2005. Samples are restricted to the eight German states included in the analysis. Samples in Panel A
contain 4,292 ethnic Germans, 9,469 other immigrants, and 320,328 native Germans, respectively. The number of
observations are lower in the other panels, depending on the sample restrictions used. Statistics are weighted by population
weights. The sample in Panel B is restricted to persons not in school. Samples in Panel C are restricted either to males of
working age, that is, 25 to 60 years old, not in school, and either employed or unemployed. Employment ratio gives the
share of employed over the male working age population. For full-time and temporary working, the sample is restricted to
employed persons only. Usual hours worked includes only full-time employed males. The sample in Panel D is restricted to
employed males of working age who are not in school. Net income is the net income in the previous month in current (2008)
Euro. Samples in Panel E include males of working age who are not in school. Hartz IV refers to public social assistance
for medium- and long-term unemployed and ALG I refers to public social assistance for short-term unemployed (< 1 year).
Standard deviations in parentheses and standard errors in brackets. Significance levels: *** p<0.01, ** p<0.05, * p<0.1.
Data source: Special evaluation of the German Microcensus 2008 by the German Federal Statistical Office.
Table A-3: Analysis Sample
State
Counties
Baden-W¨
urttemberg
Bremena
Hamburgb
Hessec
Lower Saxonyd
North Rhine-Westphaliae
Saarland
Schleswig-Holstein
44
1
1
26
46
46
6
15
Total
185
Years
1997-2006
1997-2006
1999-2006
2003-2006
1997-2006
1997-2006
1997-2006
1997-2006
County–year observations by crime category
(1)
(2)
(3)
(4)
(5)
(6)
440
10
7
104
456
460
60
150
440
3
7
104
456
138
60
150
440
10
7
104
456
460
60
150
440
3
7
104
456
138
60
150
440
10
7
104
456
138
60
150
440
3
7
104
456
138
60
150
1,687
1,358
1,687
1,358
1,365
1,358
Notes: Crime categories: (1) total crime, (2) battery, (3) drug, (4) property damage, (5) burglary, and
(6) street crime. a Bremen consists of two counties, but crime data are available only at the state level.
Crime rates for (2), (4), and (6) are available only for years 2003–2006. b Crime data are available only
for years 1999–2006. c Allocation of ethnic Germans started only in 2002. d Ethnic German inflow data
not available for city of Hannover for 2003–2006. e Crime rates for (2), (4), (5), and (6) are available
only for years 2003–2006.
Table A-4: Effect of Ethnic German Immigrants on Total Crime
Controlling for Political Majority
(1)
(2)
(3)
(4)
Dependent variable: log total crime ratet
Ethnic German inflow ratet−1
8.756**
(4.196)
Social Democrats simple majorityt−1
8.850**
(4.269)
–0.014
(0.012)
Different simple majority in state and countyt−1
8.661**
(4.281)
–0.015
(0.012)
0.008
(0.010)
Conservatives/Liberals absolute majorityt−1
0.004
(0.013)
–0.016
(0.020)
Social Democrats/Greens absolute majorityt−1
Control variables
County fixed effects
Year-by-state fixed effects
R-squared
Observations
8.969**
(4.177)
x
x
x
x
x
x
x
x
x
x
x
x
0.916
1,687
0.916
1,687
0.916
1,687
0.916
1,687
Notes: Ethnic German inflow ratet−1 is defined as the number of allocated ethnic German immigrants
in year t − 1 divided by the population in year t − 2. Control variables: log populationt−1 , log GDP
per capitat−1 , unemployment ratet−1 , foreign population sharet−1 , young male population sharet−1 ,
and clear-up ratet−1 . Social Democrats simple majorityt−1 equals 1 if the Social Democratic Party
(SPD) has the largest share of votes; 0 otherwise. Conservatives/Liberals absolute majorityt−1 equals
1 if the conservative party, the Christian Democratic Union (CDU), and the liberal party, the Free
Democratic Party (FDP), together have the majority of votes; 0 otherwise. Social Democrats/Greens
absolute majorityt−1 equals 1 if the Social Democratic Party (SPD) and the Green party together have
the majority of votes; 0 otherwise. Different majority in state and countyt−1 equals 1 if the party
with the largest share of votes in the county is different than the party with the largest share of votes
in the state; 0 otherwise. Robust standard errors in parentheses are two-way clustered at the county
and year-by-state level. Data on party votes come from the Federal Statistical Office and the State
Statistical Offices (Regionalstatistik). Shares of votes refer to second votes in the regional elections and
in the state elections, respectively. Significance levels: *** p<0.01, ** p<0.05, * p<0.1.
Table A-5: Effect of Ethnic German Immigrants on Total Crime
Including Spatially Lagged Crime Rates
(1)
(2)
(3)
Dependent variable: log total crime ratet
Ethnic German inflow ratet−1
8.756**
(4.196)
Spatial lag (distance in kilometers)
9.269**
(4.709)
–0.381
(0.813)
Spatial lag (travel distance in minutes)
Control variables
County fixed effects
Year-by-state fixed effects
R-squared
Observations
10.584**
(5.266)
–2.417
(1.665)
x
x
x
x
x
x
x
x
x
0.916
1,687
0.916
1,687
0.917
1,687
Notes: The spatial lag is the log of the sum of the total crime rate in all other counties, weighted by
the inverse of the Euclidean geographic distance between county centroids in kilometers or the travel
time distance between county capitals in car minutes. Control variables: log populationt−1 , log GDP
per capitat−1 , unemployment ratet−1 , foreign population sharet−1 , young male population sharet−1 ,
and clear-up rate for total crimet−1 . Ethnic German inflow ratet−1 equals the ethnic German inflow in
year t − 1 over population in year t − 2. Robust standard errors in parentheses are two-way clustered
at the county and year-by-state level. Significance levels: *** p<0.01, ** p<0.05, * p<0.1.